out of memory problem

[siscazzz]

Dear Seismosoft,
I have a modelling with statistic ad below :

Dynamic Time History Analysis

No. of Materials: 171
No. of Sections: 170
No. of Element Classes: 170
No. of Nodes: 2180
No. of Elements: 4030
No. of Loads: 60

When I run the model, Seismostruct give an error message : "Out of Memory".

FYI, My PC spesification :
Intel(R) Core(TM)
i5-2300 CPU @ 2.80GHz
2.79 GHz, 1.91 GB of RAM


Is it because the modelling is too large? And how can I fix this problem?

Thank you for your attention. Please reply as soon as possible

[seismosoft]

Dear siscazzz,

If you wish to run an analysis on your model as it currently stands, then you need to use a computer with more memory.

Alternatively, or in tandem, you should consider reducing the size of your model (also to reduce the duration of the analyses), since it does appear that your model is peculiarly large:
- do you really have 171 different materials in your building? We would not expect more than 3-5 different materials in a single building..
- do you also have 170 different cross-sections? Again, this seems very peculiar, since one would expect a number not exceeding 20 or so..
- are you using member subdivision? If so, perhaps you should try using one single infrmFB for each column member..

Regards,

Seismosoft Support

[siscazzz]

Dear Seismosoft,
Thankyou for your quick response
I already add 2GB RAM on my CPU, but it doesn't solve the problem. Maybe I should add more RAM?
I'm running the smaller model right now and I face another problem. I'm confused with skyline solver and frontal solver. Mine is using skyline, but my friend's model is using frontal solver. I've read "Help" section about the difference between skyline and frontal, but I cant understand which solver is better and accurate.

Mine (skyline) is slower and having more iterations than my friend's (frontal). I wonder if I used the frontal method, the process should be faster, but I dont know which one is more accurate.

Thankyou for your answer

[seismosoft]

- yes, increasing your RAM from 2GB to 4GB may allow you to run those large models that are currently leading to out-of-memory messages. (we note that if you are running a 64-bit Windows system, then adding physical RAM will suffice. Instead, if you are using a 32-bit Windows system, then, in addition to adding physical RAM, you will need also to introduce the tweaks described e.g. in http://wiki.x-plane.com/Running_Out_of_Memory)

- there should be no accuracy difference between the two Solvers. As explained in the Help System, the Frontal solver is faster, but may be less numerically stable, hence the reason for it not to be the default option. If your analysis runs fine with the Frontal solver, then there is no reason for you not to use it.

- in any case, as we repeatedly note in this Forum, if one has doubts about the results obtained with different modelling options (in this case, with the different Solvers), one needs only to run analyses with the different options, compare the results, and draw the necessary conclusions.

Seismosoft Support

[siscazzz]

Thankyou Seismosoft for the answer I think SeismoStruct has limitation for nodes (maximum 3.000 nodes for 8GB of RAM). Because mine (2.950 nodes and 5.000++ elements) finally can be run with 8GB of RAM. Mine is still running now, and maybe tomorrow I get the result. Thankyou..

[seismosoft]

No, there is no limitation of nodes in terms of RAM. There is simply a RAM limitation of 4GB in 64-bit Windows systems and 3GB in 32-bit Windows systems, as discussed in many websites, such as:
- http://msdn.microsoft.com/en-us/library/aa366521.aspx
- http://roadbiker.student.utwente.nl/Mem ... ows_OS.doc
- http://wiki.x-plane.com/Running_Out_of_Memory

Seismosoft Support